The basic problem is that the hindcast is to match the entire temperature record, not a 30 year period. The overall match is not bad, in line with what you expect, though I suspect you have a different idea of what is a good match. Many paleoclimate reconstructions have correlations of .3,.1 or even .01.

One source is IPCC AR4 Fig 9.5A,

http://bobtisdale.wordpress.com/2011/12/...

I... skept. Heavily. I'd need to see a source, and that source had better be an actual science journal or the like, or at least a reputable popular-science magazine or website (like, say, National Geographic, or Popular Science)

I could see scientists using, as least-bad, a model where the hind-casts were, say, 10% off for some particular period of recent history (while, of course, still looking for ways to improve that model). But I think if any model had hindcasts that were *that* far off, they'd scrap it and go back to the drawing board.

In fact, if hind casts are wrong, it means they overestimated or underestimated some influence on climate. They fix their error and re-run the simulation.

Why not ask the person making the claim?

This claim was made a few hours ago, though no source was provided. Now I would be extremely surprised if this was true. I don't know for certain, but I would have thought climate models would be calibrated against the record of observations, so you would expect them to get hindcasts pretty much spot on!

Does anyone have a source for the claim?

Or is it just another straw man question?